In order to reliably deliver time-sensitive media content in large scale using Internet Protocol, service providers often over-engineer a network to compensate for delays in network throughput. This can result in wasted resources in procuring the network elements and bandwidth.
Service providers can program the destination of the network element statistically relying on historical data or use load-balancing to distribute the load evenly to all elements in the pool in round-robin fashion. These techniques do not always provide for efficient use of the network elements, particularly where the media content demands are rapidly changing or where a particular cluster of servers is already overloaded. In distributed computing environments, the main distributor can provide jobs to idle servers, but in Internet Protocol Television System applications (e.g., video-on-demand), there may not be any idle servers.